Professional Documents
Culture Documents
Peter Fader
Frances and Pei-Yuan Chia Professor of Marketing
Co-Director, Wharton Customer Analytics Initiative
The Wharton School of the University of Pennsylvania
http://wcai.wharton.upenn.edu
Twitter: @faderp
WHARTON ONLINE
Challenge:
Focal
donors:
Reference:
WHARTON ONLINE
Predic/ve Analy/cs
ID
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
100001
100002
100003
100004
100005
100006
100007
100008
100009
100010
111102
111103
111104
...
WHARTON ONLINE
Predic/ve Analy/cs
L E T S F I R S T L O O K AT B O B : !
W H AT C A N W E P R E D I C T A B O U T H I S G I V I N G I N 2 0 0 2 - 0 6 ? !
ID
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
100001
100002
100003
100004
100005
100006
100007
BOB
100009
100010
...
WHARTON ONLINE
Predic/ve Analy/cs
W H AT C A N W E T E L L A B O U T S A R A H ? !
ID
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
SARAH
100002
100003
100004
100005
100006
100007
BOB
111102
111103
111104
...
WHARTON ONLINE
Predic/ve Analy/cs
ID
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
SARAH
100002
100003
MARY
100005
100006
100007
BOB
SHARMILA 1
100010
...
WHARTON ONLINE
Predic/ve Analy/cs
W H AT D O N AT I O N B E H AV I O R C H A R A C T E R I S T I C S !
D O W E N E E D T O TA K E I N T O A C C O U N T ?
G I V I N G B E H AV I O R S
D O N O R T Y P E S
R E C E N C Y:
ALIVE:
D O R M A N T:
LAPSED:
MARY
SHARMILA 1
WHARTON ONLINE
Predic/ve Analy/cs
What
does
it
mean
when
theres
one
or
more
no
dona(on
at
the
end
of
a
sequence?
a) The
donor
lapsed
(i.e.,
le`
the
donor
pool)
b) The
donor
is
dormant
(i.e.,
decided
not
to
give
that
year,
didnt
think
of
giving,
etc.)
c) We
dont
know,
but
can
build
a
model
to
come
up
with
a
best
guess
Answer:
c)
We
never
know
for
sure
whether
the
donor
is
lapsed
or
not;
based
on
recency
and
frequency
of
his
dona/on,
we
can
make
an
educated
guess
about
the
probability
of
lapsing,
so
we
can
decide
where
to
devote
resources
Based
on
our
best
guesses
about
the
probability
of
death
and
propensity
to
donate,
we
can
calculate
expected
frequency
of
future
dona(ons
for
each
donor
WHARTON ONLINE
Predic/ve Analy/cs
W H AT A B O U T M A RY V E R S U S C H R I S ? !
ID
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
SARAH
100002
100003
MARY
100005
100006
100007
BOB
111102
CHRIS
111104
...
WHARTON ONLINE
Predic/ve Analy/cs
WHARTON ONLINE
Predic/ve Analy/cs
EXCEL IMPLEMENTATION
WHARTON ONLINE
Predic/ve Analy/cs
E X P E C T E D # O F D O N AT I O N S I N 2 0 0 2 - 2 0 0 6 !
AS A FUNCTION OF RECENCY AND FREQUENCY
Name
1995
1996
1997
1998
1999
2000
2001
BOB
SARAH
MARY
SHARMILA
CHRIS
#
Rpt
Trans
(1996-2001)
1995
0.07
1
2
3
4
5
6
WHARTON ONLINE
1997
1998
1999
2000
2001
0.09
0.31
0.59
0.84
1.02
1.15
0.12
0.54
1.06
1.44
1.67
0.22
1.03
1.80
2.19
0.58
2.03
2.71
1.81
3.23
3.75
Predic/ve Analy/cs
ANALYSIS
Bob
(R:6,
F:6)
is
expected
to
donate
3.75
/mes
out
of
5
opportuni/es
between
2002
and
2006,
surprisingly
low
given
his
100%
dona/on
rate
Mary
and
Chris
have
the
same
RF
(6,4),
so
their
expected
number
of
dona/ons
going
forward
is
the
same
Even
though
Mary
and
Chris
have
lower
F
than
Sharmila,
their
higher
R
suggests
that
they
are
Alive,
thus
they
are
50%
more
valuable
than
Sharmila
Sharmila
(5,5),
despite
high
dona/on
rate,
has
likely
lapsed
Sarah,
with
very
low
R
and
F,
is
lapsed
and/or
a
very
light
donor
(hard
to
tell)
WHARTON ONLINE
#
Rpt
Trans
(1996-2001)
0
1
2
3
4
5
6
1996
1997
1998
1999
2000
2001
0.09
0.31
0.59
0.84
1.02
1.15
0.12
0.54
1.06
1.44
1.67
0.22
1.03
1.80
2.19
0.58
2.03
2.71
1.81
3.23
0.07
3.75
Predic/ve Analy/cs
WHARTON ONLINE
Predic/ve Analy/cs
WHARTON ONLINE
Predic/ve Analy/cs
A S E C O N D I L L U S T R AT I O N
Using
a
larger
dataset
from
a
dierent
non-prot
rm,
we
create
a
heat
map
that
shows
which
combina/ons
of
RF
will
likely
yield
the
most
valuable
donors
WHARTON ONLINE
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
x
/
tx
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
0
0.01
1
0.00
0.01
0.01
0.02
0.03
0.04
0.06
0.07
0.09
0.11
0.13
0.15
0.17
0.19
0.22
0.24
0.26
0.29
0.31
2
0.00
0.00
0.01
0.02
0.03
0.05
0.07
0.09
0.13
0.16
0.20
0.25
0.29
0.34
0.39
0.43
0.47
0.51
3
0.00
0.00
0.01
0.01
0.03
0.05
0.07
0.11
0.16
0.22
0.29
0.37
0.44
0.52
0.59
0.66
0.71
4
0.00
0.00
0.00
0.01
0.02
0.05
0.08
0.13
0.21
0.30
0.41
0.52
0.64
0.74
0.84
0.91
5
0.00
0.00
0.00
0.01
0.02
0.05
0.09
0.17
0.28
0.41
0.57
0.74
0.89
1.01
1.12
6
0.00
0.00
0.00
0.01
0.02
0.06
0.12
0.23
0.39
0.59
0.81
1.02
1.19
1.32
7
0.00
0.00
0.00
0.01
0.03
0.08
0.17
0.33
0.58
0.86
1.14
1.36
1.52
8
0.00
0.00
0.00
0.01
0.04
0.11
0.26
0.52
0.88
1.23
1.52
1.72
9
0.00
0.00
0.00
0.02
0.06
0.19
0.44
0.85
1.31
1.68
1.92
10
0.00
0.00
0.01
0.03
0.12
0.34
0.79
1.36
1.83
2.12
11
0.00
0.00
0.01
0.06
0.24
0.68
1.37
1.97
2.32
12
0.00
0.00
0.03
0.15
0.54
1.33
2.10
2.53
13
0.00
0.01
0.07
0.39
1.23
2.20
2.73
14
0.00
0.03
0.23
1.06
2.28
2.93
15
0.01
0.11
0.81
2.30
3.13
16
0.03
0.51
2.25
3.33
17
0.20
2.04
3.53
18
1.48
3.73
19
3.94
Predic/ve Analy/cs
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
0
0.06
0.01 0.03 0.05 0.07 0.10 0.13 0.17 0.22 0.27 0.32 0.38 0.45 0.51 0.58 0.65 0.73 0.80 0.87 0.93
0.00 0.01 0.02 0.03 0.06 0.09 0.12 0.17 0.23 0.30 0.37 0.45 0.54 0.62 0.70 0.79 0.86 0.93
0.00 0.00 0.01 0.02 0.03 0.06 0.10 0.15 0.21 0.29 0.38 0.48 0.58 0.68 0.77 0.86 0.93
0.00 0.00 0.00 0.01 0.02 0.05 0.08 0.14 0.21 0.31 0.42 0.53 0.65 0.76 0.85 0.93
0.00 0.00 0.00 0.01 0.02 0.04 0.08 0.14 0.23 0.35 0.48 0.62 0.74 0.85 0.93
0.00 0.00 0.00 0.01 0.02 0.04 0.09 0.16 0.28 0.42 0.58 0.72 0.84 0.93
0.00 0.00 0.00 0.01 0.02 0.05 0.10 0.21 0.35 0.53 0.70 0.84 0.93
0.00 0.00 0.00 0.01 0.02 0.06 0.14 0.28 0.48 0.67 0.83 0.93
0.00 0.00 0.00 0.01 0.03 0.09 0.22 0.41 0.64 0.82 0.93
10
0.00 0.00 0.00 0.01 0.05 0.15 0.35 0.60 0.81 0.93
11
12
13
14
15
16
17
18
0.37 0.93
19
0.93
WHARTON ONLINE
Predic/ve Analy/cs
Predic/ve Analy/cs
WA N T M O R E ?
For
large-scale
databases,
use
our
open-source
BTYD
R
Library:
h=p://cran.r-project.org/web/packages/BTYD/BTYD.pdf
While
this
model
oers
accurate
predic/ons
and
useful
insights
about
how
to
understand
dona/on
propensi/es,
it
stops
short
of
oering
any
specic
advice
about
which
donors
to
target
and
when
/
how
to
do
so
Building
on
this
model,
Schweidel
and
Knox
explore
the
impact
of
fundraising
eorts
on
dona/on
ac/vity
in:
Incorpora/ng
Direct
Marke/ng
Ac/vity
into
Latent
A=ri/on
Models
(h=p://ssrn.com/abstract=1670060)
WHARTON ONLINE
Predic/ve Analy/cs
DISCUSSION
Professor
Peter
Fader
www.petefader.com
Twi=er:
@faderp
Wharton
Customer
Analy/cs
Ini/a/ve
www.wharton.upenn.edu/wcai/
WHARTON ONLINE
Predic/ve Analy/cs